Fix ship/underway instrument sampling trajectories

[j-atkins]

Following #298 and the issues identified in class, this PR fixes the the path that the ship takes (and therefore the measurement locations for ADCP and Underwater_ST instruments) to be the actual shortest path between waypoints on the spherical globe.

As a reminder (and summarised in the plot below), the ship was 'missing' waypoints previously by going straight along a bearing, without curving, and this effect was enhanced when there are larger gaps between waypoints. Hence, the ship could overshoot a waypoint and leave the fieldset domain, causing out of bounds errors. Now, the logic is changed so that first an array of all locations on the forward propagation on the globe are accessed by geod.fwd_intermediate() and the number of spacings between each waypoint is determined by the distance and time between waypoints and the period associated with ADCP/Underwater_ST sampling. These are then added to the list storing where to take measurements.

Summary:

Closes #298

[erikvansebille]

Code looks good, but would be useful to also add a test that failed under the old code and passes now. In general, I think that we should adopt a more test-driven development flow. In general, every change/fix would need to be tested, to make sure there's no regression?

[j-atkins]

Test now also added to check that the ship path/underway ADCP measurement sites sit within the domain bounds determined by the waypoints and that the trajectory matches the waypoints.

[erikvansebille]

Thanks for adding the tests; these look good! I assume you tested that the old code fails on them?

[j-atkins]

I assume you tested that the old code fails on them?

Oops yes I should have indeed mentioned. The old code fails and the new passes 👍